The SQL notebook
that wires itself.
Name a cell. Any downstream cell can reference it like a table. Studio resolves the dependency graph, wires the CTEs, and runs the query: DuckDB in the browser, your warehouse, or both in a single federated join. Results inline. No install, no CLI.
What you get
Cells are models
Name a cell. Reference it from any downstream cell by name, like a table. Studio
resolves the full dependency graph and wraps each upstream output as a CTE on every
run. Reference cells across notebooks to build a shared library of transforms. No
manual wiring. No copy-paste.
Millisecond feedback
DuckDB runs as WebAssembly in your browser. Your files never leave your machine.
Connect your warehouse for production queries, or work entirely offline. Either
way, results appear under the cell in milliseconds.
Any warehouse, any source
Upload a CSV, connect Postgres, write one query that joins them. Studio routes
each segment to the right engine and returns a single result: no ETL, no staging
tables. Schema-aware completions, read-only enforced.
Interactive dashboards
Turn any notebook into a report. Drop filter controls (dropdowns, date ranges,
button groups) that wire directly to queries via
${paramName}.
Text blocks pull live values with {query.column} interpolation.
Set auto-refresh to 30 s, 1 min, or 5 min. No BI tool.Every column, profiled instantly
Click profile on any result. Get null %, unique counts, min/max, quartiles
(p25/p50/p75), mean, stddev, and top values for every column, in one pass. No
COUNT DISTINCT queries, no separate EDA step.
20+ chart types, inline
Bar, line, area, scatter, bubble, pie, histogram, heatmap, calendar heatmap,
funnel, box plot, Sankey. Each with X/Y/color/size column assignment, series mode
(grouped or stacked), and custom titles. Rendered under the cell, no tab switch.
Visual pipeline editor
Drag Filter, Group, Sort, Join, Window stages onto the canvas. PRQL updates live.
Run the pipeline up to any intermediate stage to inspect mid-transform results. Flip
to code and keep editing.
Scheduled materialization
Enable a schedule on any cell: 1 minute to 24 hours. Studio runs the cell and its
upstream dependencies, then writes the result as a table or view to your warehouse.
Cascade mode runs the full downstream graph automatically.
Generate queries with AI
Describe what you want in plain text. Studio suggests a multi-stage PRQL pipeline
with join paths, groupings, and aggregate functions inferred from your schema. Works
with any OpenAI-compatible endpoint or Ollama locally.
dbt-compatible
Open a project folder to compile, run, and test models. See the full lineage DAG:
zoom, pan, click to jump to any model. Stream test results inline without leaving
the notebook.
How it works
01
Write SQL in a cell
Each cell is a model with an output name. Downstream cells reference upstream ones
by name — no manual CTE wiring.
02
See results immediately
Hit run. DuckDB executes locally, or the query goes to your warehouse — including
cross-source joins without moving data. Results appear inline in under a second.
03
Build up a model graph
Chain cells into a full transformation pipeline. Studio assembles the WITH chain and
runs only what changed.
04
Connect to your stack
Open a project folder to compile, run, and test models from the same interface.
Works with dbt, or use Studio standalone — connect once and run SQL against any
source without ETL.
Stop waiting on the CLI.
No install required. Upload a file or connect your warehouse and start querying in seconds. Build transformation pipelines, visualize with 20+ chart types, schedule materializations, and ship to dbt, all from the browser.